Hello there Mingching, : hostA: 192.168.0.1/24 : : routerB: 192.168.0.2/24 : 192.228.118.2/24 : : routeC: 192.168.0.3/24 : 192.228.110.3/24 : : Given the hostA is rather dumb and that we can only configure one : default route, and that we have defaulted the route to B. In order : for A to be able to access C's external network, we configure : a U-turn route at B, ie the packet hop onto B and then re-forwarded : on the same interface to C. : : Is this something commonly done ? Any issue with it ? There is no problem with this. This is not uncommon--typically, router B will generate an ICMP redirect bound for host A, causing host A (if it accepts redirects) to create a route (cache) entry for the destination. If you wish traffic to move through routerB at all times, you can suppress and/or enable the generation of redirects with the sysctl net/ipv4/conf/$DEV/send_redirects [1] toggle. If host A is a linux box, you can also see if it will net/ipv4/conf/$DEV/accept_redirects I have occasionally seen peculiar TCP resets as a result of ICMP redirects not handled correctly be machines in the position of host A, but it causes no problem for routers and should pose no problem for end hosts. -Martin [1] http://ipsysctl-tutorial.frozentux.net/ipsysctl-tutorial.html#AEN630 [2] http://ipsysctl-tutorial.frozentux.net/ipsysctl-tutorial.html#AEN574 -- Martin A. Brown --- SecurePipe, Inc. --- mabrown@xxxxxxxxxxxxxx
